cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Visitor
Visitor
596 Views
Registered: ‎05-28-2019

Decent Prunning Tool

Jump to solution

Hi,

I would like to use the DECENT tool for the prunning. I have noticed in the tutorials that, it had been using as deephi_compress. Is there any complete guide how to obtain/use it?

Thanks,

Erhan

0 Kudos
Reply
1 Solution

Accepted Solutions
Xilinx Employee
Xilinx Employee
533 Views
Registered: ‎05-24-2019

@ebaturay  Xilinx has a pruning tool which is a component of DNNDK.  It is correct that at this time you should contact your local Xilinx representative in order to request access.

--Quenton

View solution in original post

0 Kudos
Reply
3 Replies
Adventurer
Adventurer
555 Views
Registered: ‎07-20-2017
According to DNNDK User Guide UG1327, prunning tool is not included in the current release (v3.0). You should contact XILINX Sales Representative or FAE to know if you can get the prunning tool.
Regards, abhidan@logictronix.com
Please mark the Answer as "Accept as solution" if information provided solves your query. Give Kudos if you think it was helpful and reply oriented.
0 Kudos
Reply
Voyager
Voyager
545 Views
Registered: ‎08-16-2018

 

I will assume you mean 'pruning' and not in the horticultural sense.

'Decent' pruning makes me wonder if there is an 'indecent' type... Do you mean effective, maybe?

If we are still on the same area, looking for an effective method for neural network pruning, TF has its optimization Kit, so you might do it on the host. Besides that, there are different approaches and algorithms for pruning so eventually you could deploy your own. NNs are highly empirical and driven by results, so actually an effective (or decent) method for you doesn't have to be that being effective/ decent for someone else working with different data and model.

 

0 Kudos
Reply
Xilinx Employee
Xilinx Employee
534 Views
Registered: ‎05-24-2019

@ebaturay  Xilinx has a pruning tool which is a component of DNNDK.  It is correct that at this time you should contact your local Xilinx representative in order to request access.

--Quenton

View solution in original post

0 Kudos
Reply